In this paper, we propose a hand posture recognition method based on the Hidden Markov Models. First, hand area is segmented from the input image using the difference images and HSV color thresholding. Next, finger points are extracted by countour tra...
In this paper, we propose a hand posture recognition method based on the Hidden Markov Models. First, hand area is segmented from the input image using the difference images and HSV color thresholding. Next, finger points are extracted by countour tracing and the distances between the finger points are calculated and modeled. Finally, using the HMM forward algorithm the user's hand posture model is compared with the preconstructed sample models. Experimental results show our method can recognize individualized hand postures as well as various types of hand postures with about 90% recognition rate. Our method can be used in game interface or sign language recognition.